Lasers are widely employed in a variety of R & D operations including spectroscopic and biotech study and industrial operations including inspecting, processing, and micromachining a variety of electronic materials and substrates. For example, to repair a dynamic random access memory (“DRAM”), laser pulses are used to sever electrically conductive links to disconnect faulty memory cells from a DRAM device, and then to activate redundant memory cells to replace the faulty memory cells. Because faulty memory cells needing link removals are randomly located, the links that need to be severed are located randomly. Thus, during the laser link repairing process, the laser pulses are fired at random pulse intervals. In other words, the laser pulses are running at a wide variable range of pulse repetition frequencies (“PRF”s), rather than at a constant PRF. For industrial processes to achieve greater production throughput, the laser pulse is fired at the target link without stopping the laser beam scanning mechanism. This production technique is referred to in the industry as “on-the-fly” (“OTF”) link processing. Other common laser applications employ laser pulses that are fired only when they are needed at random time moments.
However, the laser energy per pulse typically decreases with increasing PRF while laser pulse width increases with increasing PRF, characteristics that are particularly true for Q-switched, solid-state lasers. While many laser applications require randomly time-displaced laser pulses on demand, these applications also require that the laser energy per pulse and the pulse width be kept substantially constant. For link processing on memory or other IC chips, inadequate laser energy will result in incomplete link severing, while too much laser energy will cause unacceptable damage to the passivation structure or the silicon substrate. The acceptable range of laser pulse energies is often referred to as a “process window.” For many practical IC devices, the process window requires that laser pulse energy vary by less than 5% from a selected pulse energy value.
Skilled persons have taken various approaches for ensuring operation within a process window or for expanding the process window. For example, U.S. Pat. No. 5,590,141 for METHOD AND APPARATUS FOR GENERATING AND EMPLOYING A HIGH DENSITY OF EXCITED IONS IN A LASANT, which is assigned to the assignee of this patent application, describes solid-state lasers having lasants exhibiting a reduced pulse energy drop off as a function of PRF and, therefore, a higher usable PRF. Such lasers are, therefore, capable of generating more stable pulse energy levels when operated below their maximum PRF. U.S. Pat. No. 5,265,114 for SYSTEM AND METHOD FOR SELECTIVELY LASER PROCESSING A TARGET STRUCTURE OF ONE OR MORE MATERIALS OF A MULTIMATERIAL, MULTILAYER DEVICE, which is also assigned to the assignee of this patent application, describes using a longer laser wavelength such as 1,320 nanometers (“nm”) to expand the link process window to permit a wider variation of the laser pulse energy during the process. U.S. Pat. No. 5,226,051 for LASER PUMP CONTROL FOR OUTPUT POWER STABILIZATION describes a technique of equalizing the laser pulse energy by controlling the current of the pumping diodes. The technique works well in practical applications employing a laser PRF below about 25 KHz or 30 KHz.
The above-described laser processing applications typically employ infrared (“IR”) lasers having wavelengths from 1,047 nm to 1,342 nm, running at a PRF not over about 25 or 30 KHz. However, production needs are demanding much higher throughput, so lasers should be capable of operating at PRFs much higher than about 25 KHz, such as 50-60 KHz or higher. In addition, many laser processing applications are improved by employing ultraviolet (“UV”) energy wavelengths, which are typically less than about 400 nm. Such UV wavelengths may be generated by subjecting an IR laser to a harmonic generation process that stimulates the second, third, or fourth harmonics of the IR laser. Unfortunately, due to the nature of the harmonic generation, the pulse-to-pulse energy levels of such UV lasers are particularly sensitive to time variations in PRF and laser pulse interval.
U.S. Pat. No. 6,172,325 for LASER PROCESSING POWER OUTPUT STABILIZATION APPARATUS AND METHOD EMPLOYING PROCESSING POSITION FEEDBACK, which is also assigned to the assignee of this patent application, describes a technique of operating the laser at a constant high repetition rate in conjunction with a position feedback-controlled laser pulse picking or gating device to provide laser pulse picking on demand, at random time interval that is a multiple of the laser pulse interval, with good laser pulse energy stability and high throughput.
Typical laser pulse picking or gating devices include an acousto-optic modulator (“AOM”) and an electro-optic modulator (“EOM”), also referred to as a Pockets cell. Typical EOM material such as KD*P or KDP suffers from relatively strong absorption at the UV wavelengths, which results in a lower damage threshold of the material at the wavelength used and local heating of optical devices positioned along the laser beam path within the gating device and thereby causes changes in the voltage required by the modulator to effect one-half wavelength retardation. Another disadvantage of the EOM is its questionable ability to perform well at a repetition rate over 50 KHz. AOM material is, on the other hand, quite transparent to the UV of 250 nm up to the IR of 2,000 nm, which allows the AOM to perform well throughout typical laser wavelengths within the range. An AOM can also easily accommodate the desirable gating of pulses at a repetition rate of up to a few hundred KHz. One disadvantage of the AOM is its limited diffraction efficiency of about 75-90%.
FIG. 1 shows a typical prior art AOM 10 driven by a radio frequency (“RF”) driver 12 and employed for a laser pulse picking or gating application, and FIGS. 2A to 2D (collectively, FIG. 2) show corresponding prior art timing graphs for incoming laser pulses 14, AOM RF pulses 15, and AOM output pulses 16 and 20. FIG. 2A shows constant repetition rate laser pulses 14a-14k that are emitted by a laser (not shown) and propagated to AOM 10. FIG. 2B demonstrates two exemplary schemes for applying RF pulses 15 to AOM 10 to select which ones of laser pulses 14a-14k, occurring at corresponding time periods 22a-22k, are propagated toward a target. In a first scheme, a single RF pulse 15cde (shown in dashed lines) is extended to cover time periods 22c-22e corresponding to laser pulses 14c, 14d, and 14e; and, in a second scheme, separated RF pulses 15c, 15d, and 15e are generated to individually cover the respective time periods 22c, 22d, and 22e for laser pulses 14c, 14d, and 14e. FIGS. 2C and 2D show the respective first order beam 20 and zero order beam 16 propagated from AOM 10, as determined by the presence or absence of RF pulses 15 applied to AOM 10.
Referring to FIGS. 1 and 2, AOM 10 is driven by RF driver 12. When no RF pulses 15 are applied to AOM 10, incoming laser pulses 14 pass through AOM 10 substantially along their original beam path and exit as beam 16, typically referred to as the zero order beam 16. When RF pulses 15 are applied to AOM 10, part of the energy of incoming laser pulses 14 is diffracted from the path of the zero order beam 16 to a path of a first order beam 20. AOM 10 has a diffraction efficiency that is defined as the ratio of the laser energy in first order beam 20 to the laser energy in incoming laser pulses 14. Either first order beam 20 or zero order beam 16 can be used as a working beam, depending on different application considerations. For simplicity, laser pulses 14 entering AOM 10 will hereafter be referred as “laser pulses” or “laser output,” and pulses delivered to the target, because they are picked by AOM 10, will be referred to as “working laser pulses” or “working laser output.”
When the first order beam is used as the working beam, the energy of the working laser pulses can be dynamically controlled from 100% of its maximum value down to substantially zero, as the RF power changes from its maximum power to substantially zero, respectively. Because the practical limited diffraction efficiency of an AOM 10 under an allowed maximum RF power load is about 75% to 90%, the maximum energy value of the working laser pulses is about 75-90% of the laser pulse energy value from the laser. However, when the zero order beam 16 is used as the working beam, the energy of the working laser pulses can be dynamically controlled from 100% of the maximum value of the laser pulse energy from the laser down to 15-20% of the maximum value, as the RF power changes from substantially zero to its maximum power, respectively. For memory link processing, for example, when the working laser pulse is not on demand, no leakage of system laser pulse energy is permitted, i.e., the working laser pulse energy should be zero so the first order laser beam 20 is used as the working beam.
With reference again to FIG. 2, RF pulses 15 are applied to AOM 10 at random time intervals and only when working laser pulses are demanded, in this case, at random integral multiples of the laser pulse interval. The random output of working laser pulses results in random variable thermal loading on AOM 10. Variable thermal loading causes geometric distortion and temperature gradients in AOM 10, which cause gradients in its refractive index. The consequences of thermal loading distort a laser beam passing through AOM 10, resulting in deteriorated laser beam quality and instability in the laser beam path or poor beam positioning accuracy. These distortions could be corrected to some degree if they could be kept constant. However, when the system laser pulses are demanded randomly, such as in laser link processing, these distortions will have the same random nature and cannot be practically corrected.
Test results on an AOM device, such as a Model N23080-2-1.06-LTD, made by NEOS Technologies, Melbourne, Fla., showed that with only 2 W RF power, the laser beam pointing accuracy can deviate as much as 1 mrad when the RF to the AOM 10 is applied on and off randomly. This deviation is a few hundred times greater than the maximum allowed for the typical memory link processing system. Laser beam quality distortion due to the random thermal loading on the AOM 10 will also deteriorate the focusability of the laser beam, resulting in a larger laser beam spot size at the focusing point. For applications such as the memory link processing that require the laser beam spot size to be as small as possible, this distortion is very undesirable.
What is needed, therefore, is an apparatus and a method for randomly picking working laser pulses from a high repetition rate laser pulse train without causing distortion to the laser beam quality and positioning accuracy due to the random thermal loading variation on the AOM. What is also needed is an apparatus and method of generating working laser pulses having constant laser energy per pulse and constant pulse width on demand and/or on-the-fly at a high PRF and with high accuracy at vastly different pulse time intervals for a variety of laser applications such as spectroscopic, bio-tech, or micromachining applications, including laser link processing on memory chips.